Raspberry Pi 3でGoogle Cloud Visionを使う
Googleの画像認識エンジンである、Cloud VisionをRaspberry Pi 3で使ってみたので、その備忘録。参考にしたのは、以下。
- Raspberry PiでNode.jsを動かす - Qiita
- MacにNode.jsをインストールする - Qiita
- Node.jsとGoogle Cloud Vision API使って色々な画像認識試してみた。 | Indie Engineer Blog
①Node.jsをRaspberry Pi 3に導入する
まずはnvm(Node Version Manager)の導入。
$ git clone https://github.com/creationix/nvm.git ~/.nvm
次にパスを通す
$ source ~/.nvm/nvm.sh
これでnvmコマンドが使えるようになったので、Node.js公式を見て、最新バージョンのNode.jsをインストールする。最新バージョンの確認は下記nvm ls-remoteコマンドでもわかる。
$ nvm ls-remote
$ nvm install v5.10.1
次にNode.jsの使用バージョンのデフォルト設定を行う。
$ nvm alias default v5.10.1
バージョンを確認してみる。
$ node -v
v5.10.1
動いていそうだ。
.bash_profileにNode.js周りの設定を書いておく。
$ vi .bash_profile
ここを参考にした。
# nvm設定
[[ -s ~/.nvm/nvm.sh ]] && . ~/.nvm/nvm.sh
nvm use default
npm_dir=${NVM_PATH}_modules
export NODE_PATH=$npm_dir
次に以下コマンドを打ち設定を反映させる。
$ source ~/.bash_profile
これでNode.jsの準備は完了。
②Google Cloud Visionの設定
https://cloud.google.com/vision/にアクセスして、登録をすませる。
左上のハンバーガーメニューから、API Managerを選択し、新しい認証情報→APIキー
その後サーバーキーを選択して、作成。
そうそう、無料試用期間とはいえ、万が一が恐いので、請求のアラート設定をしておく。例えば、APIキーが漏れてしまった時などでも、被害を最小限にとどめられる。お支払い→予算とアラート→予算を作成から適当な金額を設定する。AWS破産とか冗談にならないからね。
③node-cloud-vision-apiをRaspberry Pi 3に導入する
適当なディレクトリに移動し、フォルダを作る
$ cd home/pi/projects/
$ mkdir cloudvision
npm initして、node-cloud-vision-apiをインストールする
$ npm init
$ npm install --save node-cloud-vision-api
④JavaScriptを実行してみる
'use strict'
const vision = require('node-cloud-vision-api')
// init with auth
// Your API Keyのところを②で取得したAPIキーに変える。
// これは公開すると他人に使われる恐れがあるので公開しないように！！
vision.init({auth: 'Your API Key'})
// construct parameters
const req1 = new vision.Request({
// ローカルにあるイメージを指定。
image: new vision.Image('/home/pi/Pictures/summerfruits.jpg'),
features: [
new vision.Feature('FACE_DETECTION', 4),
new vision.Feature('LABEL_DETECTION', 10),
]
})
// 2nd image of request is load from Web
const req2 = new vision.Request({
image: new vision.Image({
// インターネット上のイメージを指定。
url: 'https://scontent-nrt1-1.cdninstagram.com/hphotos-xap1/t51.2885-15/e35/12353236_1220803437936662_68557852_n.jpg'
}),
features: [
new vision.Feature('FACE_DETECTION', 1),
new vision.Feature('LABEL_DETECTION', 10),
]
})
// send multi requests by one API call
vision.annotate([req1, req2]).then((res) => {
// handling response for each request
console.log(JSON.stringify(res.responses))
}, (e) => {
console.log('Error: ', e)
})
上記コードをcloudvision_samplecode.jsとして保存する。
$ node cloudvision_samplecode.js
として実行すると、下記がターミナルに出力される。（見やすくするためにインデントをつけている）
{
"labelAnnotations":
[
{
"mid":"/m/036qh8",
"description":"produce",
"score":0.97685134
},
{
"mid":"/m/02wbm",
"description":"food",
"score":0.94573361
},
{
"mid":"/m/05s2s",
"description":"plant",
"score":0.94254869
},
{
"mid":"/m/02xwb",
"description":"fruit",
"score":0.90469909
},
{
"mid":"/m/0dxb5",
"description":"berry",
"score":0.88049006
},
{
"mid":"/m/0gqbt",
"description":"shrub",
"score":0.53203881
},
{
"mid":"/m/0463nlw",
"description":"west indian raspberry",
"score":0.50897217
}
]
},
{
"faceAnnotations":
[
{
"boundingPoly":
{
"vertices":
[
{"x":112},
{"x":724},
{"x":724,"y":510},
{"x":112,"y":510}
]
},
"fdBoundingPoly":
{
"vertices":
[
{"x":193},
{"x":638},
{"x":638,"y":399},
{"x":193,"y":399}
]
},
"landmarks":
[
{
"type":"LEFT_EYE",
"position":{"x":342.14697,"y":77.257805,"z":0.0010066124}
},
{
"type":"RIGHT_EYE",
"position":{"x":524.90961,"y":100.77812,"z":0.84962881}
},
{
"type":"LEFT_OF_LEFT_EYEBROW",
"position":{"x":285.53159,"y":30.352997,"z":25.532389}
},
{
"type":"RIGHT_OF_LEFT_EYEBROW",
"position":{"x":395.71207,"y":32.727783,"z":-25.809153}
},
{
"type":"LEFT_OF_RIGHT_EYEBROW",
"position":{"x":484.55322,"y":44.141739,"z":-25.38883}
},
{
"type":"RIGHT_OF_RIGHT_EYEBROW",
"position":{"x":591.29803,"y":69.55114,"z":26.930927}
},
{
"type":"MIDPOINT_BETWEEN_EYES",
"position":{"x":434.99573,"y":73.941544,"z":-36.170986}
},
{
"type":"NOSE_TIP",
"position":{"x":423.93985,"y":167.91109,"z":-111.36714}
},
{
"type":"UPPER_LIP",
"position":{"x":411.80557,"y":241.09036,"z":-82.791382}
},
{
"type":"LOWER_LIP",
"position":{"x":405.9841,"y":306.68048,"z":-82.744064}
},
{
"type":"MOUTH_LEFT",
"position":{"x":337.35049,"y":267.53821,"z":-39.726524}
},
{
"type":"MOUTH_RIGHT",
"position":{"x":487.37915,"y":290.97012,"z":-37.928776}
},
{
"type":"MOUTH_CENTER",
"position":{"x":410.22937,"y":273.30365,"z":-75.321}
},
{
"type":"NOSE_BOTTOM_RIGHT",
"position":{"x":471.14438,"y":202.10455,"z":-44.104683}
},
{
"type":"NOSE_BOTTOM_LEFT",
"position":{"x":371.49542,"y":186.17801,"z":-44.648823}
},
{
"type":"NOSE_BOTTOM_CENTER",
"position":{"x":419.42804,"y":201.44179,"z":-75.345909}
},
{
"type":"LEFT_EYE_TOP_BOUNDARY",
"position":{"x":343.91556,"y":62.291443,"z":-9.4704609}
},
{
"type":"LEFT_EYE_RIGHT_CORNER",
"position":{"x":378.49484,"y":83.426445,"z":0.63221854}
},
{
"type":"LEFT_EYE_BOTTOM_BOUNDARY",
"position":{"x":340.5495,"y":88.847572,"z":-4.0131183}
},
{
"type":"LEFT_EYE_LEFT_CORNER",
"position":{"x":304.99127,"y":76.559814,"z":16.381186}
},
{
"type":"LEFT_EYE_PUPIL",
"position":{"x":339.22885,"y":75.820984,"z":-4.330039}
},
{
"type":"RIGHT_EYE_TOP_BOUNDARY",
"position":{"x":526.98669,"y":85.809883,"z":-8.6181784}
},
{
"type":"RIGHT_EYE_RIGHT_CORNER",
"position":{"x":560.83966,"y":109.43663,"z":17.566591}
},
{
"type":"RIGHT_EYE_BOTTOM_BOUNDARY",
"position":{"x":523.4292,"y":112.47009,"z":-3.1689837}
},
{
"type":"RIGHT_EYE_LEFT_CORNER",
"position":{"x":485.92252,"y":100.39104,"z":0.7664215}
},
{
"type":"RIGHT_EYE_PUPIL",
"position":{"x":527.13208,"y":100.01988,"z":-3.736815}
},
{
"type":"LEFT_EYEBROW_UPPER_MIDPOINT",
"position":{"x":343.1142,"y":7.1320877,"z":-7.4442763}
},
{
"type":"RIGHT_EYEBROW_UPPER_MIDPOINT",
"position":{"x":541.75824,"y":32.65345,"z":-6.5013323}
},
{
"type":"LEFT_EAR_TRAGION",
"position":{"x":207.46692,"y":196.30023,"z":203.98279}
},
{
"type":"RIGHT_EAR_TRAGION",
"position":{"x":623.50317,"y":249.52588,"z":205.91609}
},
{
"type":"FOREHEAD_GLABELLA",
"position":{"x":440.61343,"y":34.804077,"z":-31.940067}
},
{
"type":"CHIN_GNATHION",
"position":{"x":395.04126,"y":391.9465,"z":-74.228394}
},
{
"type":"CHIN_LEFT_GONION",
"position":{"x":215.1674,"y":298.48297,"z":105.92882}
},
{
"type":"CHIN_RIGHT_GONION",
"position":{"x":590.98389,"y":346.83713,"z":107.62403}
}
],
"rollAngle":7.2499251,
"panAngle":0.27318951,
"tiltAngle":11.967749,
"detectionConfidence":0.577574,
"landmarkingConfidence":0.19900249,
"joyLikelihood":"POSSIBLE",
"sorrowLikelihood":"VERY_UNLIKELY",
"angerLikelihood":"VERY_UNLIKELY",
"surpriseLikelihood":"VERY_UNLIKELY",
"underExposedLikelihood":"VERY_UNLIKELY",
"blurredLikelihood":"VERY_UNLIKELY",
"headwearLikelihood":"VERY_UNLIKELY"}],
"labelAnnotations":
[
{"mid":"/m/01k74n","description":"facial expression","score":0.86115927},
{"mid":"/m/068hy","description":"pet","score":0.80557323},
{"mid":"/m/017ftj","description":"sunglasses","score":0.72584444},
{"mid":"/m/0sgh53y","description":"selfie","score":0.71132851}
]
}
]
画像の一枚目は下のフルーツの画像で、
{
"mid":"/m/036qh8",
"description":"produce",
"score":0.97685134
},
{
"mid":"/m/02wbm",
"description":"food",
"score":0.94573361
},
{
"mid":"/m/05s2s",
"description":"plant",
"score":0.94254869
},
{
"mid":"/m/02xwb",
"description":"fruit",
"score":0.90469909
},
{
"mid":"/m/0dxb5",
"description":"berry",
"score":0.88049006
},
{
"mid":"/m/0gqbt",
"description":"shrub",
"score":0.53203881
},
{
"mid":"/m/0463nlw",
"description":"west indian raspberry",
"score":0.50897217
とあるので、大体は合っていそう。
画像の２枚目はTaylor Swiftである。
"joyLikelihood":"POSSIBLE",
"sorrowLikelihood":"VERY_UNLIKELY",
"angerLikelihood":"VERY_UNLIKELY",
"surpriseLikelihood":"VERY_UNLIKELY",
"underExposedLikelihood":"VERY_UNLIKELY",
"blurredLikelihood":"VERY_UNLIKELY",
"headwearLikelihood":"VERY_UNLIKELY"}],
"labelAnnotations":
[
{"mid":"/m/01k74n","description":"facial expression","score":0.86115927},
{"mid":"/m/068hy","description":"pet","score":0.80557323},
{"mid":"/m/017ftj","description":"sunglasses","score":0.72584444},
{"mid":"/m/0sgh53y","description":"selfie","score":0.71132851}
]
sorrow,anger,などがunlikelyなのであっている。サングラスや自撮りであることも認識していそうである。
結構面白い。月に1000回ぐらいしか無料のAPIコール枠はないのでやりすぎないように注意。
